EFFECTS OF FOREST GROWTH ON LASER DERIVED CANOPY METRICS
T. Gobakken a,b, E. Næsset a
a
Agricultural University of Norway, Department of Ecology and Natural Resource Management, P.O. Box 5003, N1432 Ås, Norway – (terje.gobakken, erik.naesset)@ina.nlh.no
b
The Norwegian Institute of Land Inventory, The Norwegian National Forest Inventory, P.O. Box 115, N - 1431 Ås,
Norway
KEY WORDS: Airborne laser scanning, change detection, forest, growth
ABSTRACT:
In the current research, we present the results from a growth detection study in young forest and mature pine-dominated forest
located on poor sites. Canopy height distributions were created from small-footprint airborne laser scanner data collected over 87
georeferenced field sample plots. The size of each plot was 300-400 m2. The laser data were acquired in 1999 and 2001, and none of
the plots had been subject to any harvests or serious natural disturbances in the period between the two laser data acquisitions.
Canopy height distributions were created from first pulse data. Height percentiles (10%, 50% and 90%), mean and maximum height
values, and canopy density at three different height intervals above the ground were computed from the laser-derived canopy height
distributions. Corresponding metrics derived from the laser data acquired in 1999 and 2001 were compared. All metrics derived from
the first pulse data differed significantly between data acquired in 1999 and 2001 in the young forest and seven of eight metrics
differed significantly in the mature forest. The mean values for the difference in mean height and maximum height were 1.24 and
1.11 m in young and 0.43 and 0.22 m mature forest, respectively.
1. INTRODUCTION
Monitoring of timber resources, biomass and carbon stocks
might be one of several major directions of future utilization of
small-footprint laser data in forests. With an objective of
detecting possible vegetation change Sweda et al. (2003) made
two temporally separated airborne laser profiling transects
flown at five-year interval along a NS-oriented 600 km transect
across the boreal forest of western Canada. Yu et al. (2004)
were the first to demonstrate the applicability of small footprint,
high sampling density airborne laser scanners for boreal forest
change detection, i.e. the estimation of forest growth and
monitoring of harvested trees. The first change detection studies
based on laser scanner data over forest indicate a large potential
of multi-temporal data for forest monitoring.
The objective of the present study was (1) to analyse the effect
of forest growth at plot level over a short time period on laserderived metrics typically used in forest inventory and (2) to
assess how these effects are influenced by forest type.
2. MATERIALS AND METHODS
2.1 Study area
This study was based on data from a forest inventory in
southeast Norway conducted in the municipality of Våler
(59°30′N, 10°55′E, 70–120 m a.s.l.). The size of the inventory
was approximately 1000 ha. Compared to average forest
conditions in Norway, the terrain was considered as quite flat
with gentle slopes. The main tree species were Norway spruce
[Picea abies (L.) Karst.] and Scots pine (Pinus sylvestris L.).
Further details about the study area can be found in Næsset
(2002b). Sample plots were used to assess the effects of forest
growth on laser-derived canopy height and density metrics
derived from laser data acquired in 1999 to 2001.
2.2 Field data
Field data were collected during summer of 1998 (see Næsset,
2002a). Since the laser data were acquired in 1999 and 2001
(see below), the area were revisited in field in December 2001
to verify that plots had not been subject to any harvests or
serious natural disturbances. However, it is likely that some
natural mortality had occurred during the 3–4 year period.
In this study, the plots with young forest ("stratum I") and
mature forest with poor site quality ("stratum II") were
analyzed. In total, 87 circular sample plots were distributed
systematically throughout the entire 1000 ha study area
according to a regular grid. The size of each plot was 300 m2 in
stratum I and 400 m2 in stratum II.
On each plot, all trees with diameter at breast height (dbh)>4
and >10 cm were callipered on young and mature plots,
respectively, which conforms to ordinary inventory practice in
Norway. Basal area (G) was computed as the basal area per
hectare of the callipered trees. The heights of sample trees
selected with probability proportional to stem basal area at
breast height using a relascope were measured by a Vertex
hypsometer. Mean height of each plot was computed as Lorey's
mean height (hL), i.e., mean height weighted by basal area.
Total plot volume (V) was computed according to the tree
volume tariff method (Husch et al., 1993) by calculating the
sum of the individual tree volumes for trees with dbh>4 cm and
dbh>10 cm, respectively. hL, G, and V were prorated by up to
3.8 years using growth models (Blingsmo, 1984; Braastad,
1975; Braastad, 1980; Fitje and Vestjordet, 1977) to correspond
to the dates on which the 1999 and 2001 laser data were
acquired. These prorated values were used as ground reference.
A summary of the ground-truth plot data is displayed in Table
1.
- 224 -
International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XXXVI - 8/W2
1999
Characteristic
Range
Mean
Mean
Growth
2001
Young forest – stratum I (n=53)
hL (m)
6.5
-
21.2
13.6
0.9
14.5
G (m2ha-1)
9.7
-
41.4
24.5
2.9
27.4
3
36.3
-
460.1
178.6
28.6
207.2
-1
V (m ha )
Tree species distribution (%)
Spruce
0
-
100
53
Pine
Deciduous
species
0
-
97
34
0
-
69
13
Mature forest, poor site quality – stratum II (n=34)
hL (m)
11.4
-
21.5
16.1
0.2
16.4
G (m ha )
9.4
-
29.5
19.1
0.8
19.9
V (m3ha-1)
56.4
-
273.8
148.7
8.3
157.0
2
-1
Tree species distribution (%)
Spruce
0
-
89
ellipsoidic height values were computed for all first and last
returns. The last return data were used to model the ground
surface. In a filtering operation on the last return data
undertaken by the contractor using a proprietary routine, local
maxima assumed to represent vegetation hits were discarded. A
triangulated irregular network (TIN) was generated from the
planimetric coordinates and corresponding height values of the
individual terrain ground points retained in the last pulse
dataset. Individual TIN models were made of the 1999 and
2001 laser data. The ellipsoidic height accuracy of the TIN
models was expected to be around 25 cm (Kraus and Pfeifer,
1998; Reutebuch et al., 2003).
For both the 1999 and the 2001 laser data, all first return
observations (points) were spatially registered to the TIN
according to their coordinates. Terrain surface height values
were computed for each point by linear interpolation from the
TIN. The relative height of each point was computed as the
difference between the height of the first return and the terrain
surface height.
Observations with a height value less than 2 m were excluded
from the dataset to eliminate ground hits and the effect of
stones, shrubs, etc. from the tree canopy datasets (Nilsson,
1996). Pulses that hit outside these objects were excluded from
further analysis.
29
Pine
0
100
66
Deciduous
species
0
21
5
a
hL=Lorey's mean height, G=basal area, V=volume.
Table 1. Summary of sample plot reference data a
Differential GPS and Global Navigation Satellite System
(GLONASS) were used to determine the position of the centre
of each sample plot. A Javad Legacy 20-channel dualfrequency receiver observing pseudorange and carrier phase of
both GPS and GLONASS was used. Another Javad Legacy
receiver was used as base station. The distance between the
plots and the base station was approximately 19 km.
Coordinates were computed by postprocessing in an adjustment
with coordinates and carrier phase ambiguities as unknown
parameters using both pseudorange and carrier phase
observations. The estimated accuracy of the planimetric plot
coordinates (x and y) ranged from <0.1 to 2.5 m with an
average of approximately 0.3 m. Further details are given by
Næsset (2002a).
2.3 Laser scanner data
Laser scanner data for this specific study were acquired 8 and 9
June 1999, and 16 and 17 July 2001 (leaf-on canopy
conditions). A Piper PA31-310 aircraft carried the ALTM 1210
laser scanner system. The pulse repetition frequency was 10
kHz on both occasions. The average flying altitudes were
approximately 700 m a.g.l. in 1999 and 840 m a.g.l. in 2001,
and the average footprint densities were 1.1 m-2 and 0.9 m-2.
According to standard procedures, pulses that were transmitted
at scan angles that exceeded 14° and 15° in 1999 and 2001,
respectively, were excluded from the final datasets.
Processing of the laser data was accomplished by the contractor
(BN Mapping, Norway). Planimetric coordinates (x and y) and
2.4 Computations
In this study, height distributions were created from laser pulses
classified as canopy hits (>2 m, see above) for each plot
inventoried in field, and percentiles for the canopy height for
10% (h10), 50% (h50), and 90% (h90) were computed. In
addition, also the maximum (hmax) and mean values (hmean) of
the height distributions were computed. Furthermore, several
measures of canopy density were derived. The range between
the lowest laser canopy height (>2 m) and the maximum
canopy height was divided into 10 fractions of equal length.
Canopy densities were then computed as the proportions of
laser hits above fraction #0 (>2 m), 1, . . ., 9 to total number of
pulses. The densities for fraction #1 (d1), #5 (d5), and #9 (d9)
were selected for further studies.
The mean differences between data acquired in 2001 and 1999
of the investigated height and density-related metrics were
compared within different strata by means of paired t-tests to
assess how forest growth influenced on the laser-derived
metrics. The mean differences between the data acquired in
2001 and 1999 were also compared between strata to assess
how these metrics were influenced by forest type.
Correspondingly, the variances of the differences between laser
data acquired on the two occasions were compared for different
strata by means of F tests.
In the comparisons of the eight laser-derived metrics within and
between strata, eight t-tests or F tests were accomplished
simultaneously. In order to control the total Type I error,
Bonferroni tests were applied (Miller, 1981). Thus, the level of
significance for each of the eight tests was α/8.
3. RESULTS
Height percentiles: First, laser-derived canopy height and
density metrics were computed for the 87 sample plots. For the
first pulse height distribution percentiles (h10, h50, h90), all the
mean differences between data acquired in 2001 and 1999 were
found to be statistically significant. The differences ranged
- 225 -
International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XXXVI - 8/W2
from 0.40 to 1.32 m (Table 2). All of the differences were
found to be statistically significant (p>0.05) when the
comparisons between forest types were made (Table 3).
The standard deviations for the differences of the percentiles
h10, h50, h90 between laser data acquired in 2001 and 1999
ranged from 0.33 to 0.72 m for the sample plots (Table 2).
None of the variances for the differences were found to be
statistically significant (p>0.05) when the comparisons between
forest types were made (Table 3). Both the mean differences
and the standard deviations for the differences were larger in
the young forest (stratum I) as compared to the mature pine
forest on poor sites (stratum II).
Metricsb
D, first pulse
Mean
Young forest - stratum I (n= 53 sample plots)
S.D.
h10 (m)
1.00
***
0.72
h50 (m)
1.31
***
0.48
h90 (m)
1.32
***
0.46
hmax (m)
1.11
***
0.86
hmean (m)
1.24
***
0.42
d1 (%)
0.06
***
0.04
d5 (%)
0.11
***
0.05
d9 (%)
0.01
**
0.02
3). The standard deviations were 0.86 m in stratum I and 0.74 m
stratum II.
The mean height values of the first pulse data (hmean) showed
similar patterns as did the height percentiles. The mean
differences between laser data acquired in 2001 and 1999 were
1.24 m in stratum I and 0.43 m in stratum II. The differences
were found to be statistically significant (p>0.05). The
differences across data acquired in 1999 and 2001 differed
significantly between young forest and mature forest dominated
by pine (Table 3). The standard deviations for the differences in
mean values between data acquired in 2001 and 1999 were 0.42
m in stratum I and 0.29 m in stratum II. The standard deviations
for the differences in mean height values tended to be smaller
than the corresponding standard deviations for percentiles and
maximum values.
Canopy density: For the first pulse canopy densities (d1, d5 and
d9), the mean differences ranged from 0.01% to 0.11% (Tables
2). The density variables varied significantly between data
acquired in 1999 and 2001 for all comparisons. The differences
between data acquired in 2001 and 1999 differed significantly
between young forest and mature pine forest for d5 (Table 3)
The standard deviations for differences between data acquired
in 2001 and 1999 ranged between 0.02% and 0.05% for all
comparisons and were smallest for d9.
_
Mature forest, poor site quality - stratum II (n= 34 sample
plots)
h10 (m)
0.43
**
0.67
h50 (m)
0.48
***
0.39
h90 (m)
0.40
***
0.33
hmax (m)
0.22
NS
0.74
hmean (m)
0.43
***
0.29
d1 (%)
0.05
***
0.05
d5 (%)
0.05
***
0.03
d9 (%)
0.01
*
0.02
Level of significance (Bonferroni test, α/8): NS = not
significant (>0.05); *< 0.05; **< 0.01; ***< 0.001.
b
h10, h50, and h90 = percentiles of the laser canopy heights for
10%, 50%, and 90%; hmax = maximum laser canopy height;
hmean = arithmetic mean laser canopy height; d1, d5, and d9 =
canopy densities corresponding to the proportions of laser hits
above fraction # 1, 5, and 9, respectively, to total number of
pulses (see text).
a
Table 2. Differences (D) between first pulse laser scanner data
from 2001 and 1999, respectively, for laser-derived
metrics of sample plots and standard deviations
(S.D.) for the differences a
Height maximum and mean: The maximum values of first
pulse canopy height distributions (hmax) did not differ
significantly between the data acquired in 2001 and 1999 for
mature forest dominated by pine (stratum II) (Table 2). The
mean differences between data acquired in 2001 and 1999 were
1.11 m in stratum I and 0.22 m in stratum II. The differences
for the plots differ significantly between the two strata (Table
_
Metrics
D I- D II
S.D.I-S.D.IId
h10 (m)
**
NS
h50 (m)
***
NS
h90 (m)
***
NS
hmax (m)
***
NS
hmean (m)
***
NS
d1 (%)
NS
NS
d5 (%)
***
NS
d9 (%)
NS
NS
Roman subscript refers to stratum.
b
Level of significance (Bonferroni test, α/8): NS = not
significant (>0.05); *< 0.05; **< 0.01; ***< 0.001.
c
h10, h50, and h90 = percentiles of the laser canopy heights for
10%, 50%, and 90%; hmax = maximum laser canopy height;
hmean = arithmetic mean laser canopy height; d1, d5, and d9 =
canopy densities corresponding to the proportions of laser hits
above fraction # 1, 5, and 9, respectively, to total number of
pulses (see text).
d
Variance comparison by F tests.
a
_
Table 3. Comparisons between strata of mean differences ( D )
and standard deviations for the differences (S.D.)
between laser derived metrics from laser data
acquired in 2001 and 1999 for the sample plotsa,b,c
4. DISCUSSION
Detection of forest growth by means of laser scanning is
important with respect to future use of LIDAR in large-area
timber resource and carbon stock monitoring. To the very best
of our knowledge, the only study dealing with determination of
forest growth and laser scanning so far is the work by Yu et al.
(2004) where single trees were detected. Yu et al. (2004)
developed an object-oriented tree-to-tree matching algorithm.
Out of 83 field-checked harvested trees, 61 could be
- 226 -
International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XXXVI - 8/W2
automatically detected. All mature harvested trees were
detected and it was mainly the smaller trees that caused
problems. The precision of the estimated height growth, based
on field checking or statistical analysis, was about 5 cm at stand
level and 10-15 cm at plot level. However, the practical laserbased inventory procedure proposed by Næsset and Bjerknes
(Næsset and Bjerknes, 2001) and Næsset (2002b) and areabased measurement are more suitable for large-area monitoring
than tree-to-tree matching due to the large amount of data
needed for single tree measurements.
In the present study, changes at 300-400 m2 plots over a twoyear period were assessed and the major findings of this study
are that height percentiles and mean values of the first pulse
data from 2001 were higher than corresponding values derived
from 1999 data, and that even certain canopy density measures
may be suitable to detect forest growth.
The first pulse measurement of hmax was not suitable for
detecting forest growth, at least not in the slow-growing mature
pine forest. This may be due to the fact that larger sampling
errors are associated with the maximum value than, for
example, the height percentiles (Næsset, 2004). This was also
observed in the present data. The standard deviations for the
differences between maximum values were larger than
corresponding standard deviations for the other height-related
metrics.
In this study, a growth period of two years was analyzed. In e.g.
ordinary growth and yield studies or national forest inventories
the changes in five year periods are usually applied. More
research should be carried out to study how short growthperiods that can be used in such studies. If the time period is too
short – especially for low productive forest types – the growth
might not be detected due to other error sources.
In spite of the uncertainties related to the mortality and the
length of the growth period, forest growth for different forest
types can be revealed by laser-derived metrics at plot level.
However, to improve the potential of laser-scanning in growth
detection, it should be evaluated (1) how effective laserscanning is for growth detection for other forest types and (2) in
entire stands instead of small plots, and (3) the suitability of
other laser derived metrics than those considered here and (4)
of last pulse data should be investigated.
Kraus, K. and Pfeifer, N., 1998. Determination of terrain
models in wooded areas with airborne laser scanner data. ISPRS
J. Photogramm. Remote Sens., 53: 193-203.
Miller, R.G., 1981. Simultaneous statistical inference. SpringerVerlag, New York, 299 pp.
Nilsson, M., 1996. Estimation of tree heights and stand volume
using an airborne lidar system. Remote Sens. Environ., 56: 1-7.
Næsset, E., 2002a. Determination of mean tree height of forest
stands by means of digital photogrammetry. Scand. J. For.
Res., 17: 446-459.
Næsset, E., 2002b. Predicting forest stand characteristics with
airborne scanning laser using a practical two-stage procedure
and field data. Remote Sens. Environ., 80: 88-99.
Næsset, E., 2004. Effects of different flying altitudes on
biophysical stand properties estimated from canopy height and
density measured with a small-footprint airborne scanning
laser. Remote Sens. Environ., 91: 243–255.
Næsset, E. and Bjerknes, K.-O., 2001. Estimating tree heights
and number of stems in young forest stands using airborne laser
scanner data. Remote Sens. Environ., 78: 328-340.
Reutebuch, S.E., McGaughey, R.J., Andersen, H.-E. and
Carson, W.W., 2003. Accuracy of a high-resolution lidar terrain
model under a conifer forest canopy. Canadian Journal of
Remote Sensing, 29: 527-535.
Sweda, T., Tsuzuki, H. and Kusakabe, T., 2003. Climateinduced vegetation change in Canadian boreal forest as
detected by airborne laser profiling. In: J. Hyyppä, E. Næsset,
H. Olsson, T. Granquist Palen and H. Reese (Editors),
Scandlaser Scientific Workshop on Aiborne laser scanning of
forests. Working Paper 112. Swedish University of Agricultural
Sciences, Department of Forest Resource Management and
Geomatics, Umeå Sweden, pp. 108-114.
Yu, X., Hyyppä, J., Kaartinen, H. and Maltamo, M., 2004.
Automatic detection of harvested trees and determination of
forest growth using airborne laser scanning. Remote Sens.
Environ., 90(4): 451-462.
REFERENCES
Blingsmo, K.R., 1984. Diameter increment functions for stands
of birch, Scots pine and Norway spruce. Research Paper 7/84.
Norw. for. Res. Inst., 22 pp.
Braastad, H., 1975. Yield tables and growth models for Picea
abies, 31. Rep. Norw. For. Res. Inst., Ås, 364-537 pp.
Braastad, H., 1980. Growth model computer program for Pinus
sylvestris. Meddr. Norw. For. Res. Inst., 35, 272-359 pp.
Fitje, A. and Vestjordet, E., 1977. Stand height curves and new
tariff tables for Norway spruce. Rep. Norw. For. Res. Inst., 34:
23-62.
Husch, B., Miller, C.I. and Beers, T.W., 1993. Forest
mensuration. Krieger Publishing, Malabar, FL, 402 pp.
- 227 -